CCS Timing Technical White Paper
CCS Timing Technical White Paper
CCS Timing Technical White Paper
CCS Noise
Version 1.1
Abstract
This document describes the Synopsys CCS Noise model for cell-level noise
analysis.
Version 1.1
11/21/05
1
Synopsys, Inc.
700 East Middlefield Road
Mountain View, CA 94043 USA
Version 1.1
11/21/05
2
Synopsys, Inc.
700 East Middlefield Road
Mountain View, CA 94043 USA
1 Introduction
CCS Noise is a new advanced current-based driver model that enables
accurate noise analysis with results very close to SPICE simulation. It
not only precisely models injected crosstalk noise bumps, but also allows
more advanced analysis, such as propagated noise bumps and the driver
weakening, without significant characterization effort.
With CCS Noise, the noise immunity of the cells can also be obtained
during the analysis by using the actual noise bump waveforms without
the additional need for separate characterization. This dynamic
computation of noise propagation and noise immunity enables noise
library characterization to be 100 times faster than NLDM noise
characterization.
There are three main components to noise calculation: driver modeling,
receiver modeling, and reduced order modeling of parasitics. The focus
of this paper is the CCS Noise driver model; information about the
advanced receiver modeling can be found in the CCS Timing Technical
White Paper [1].
11/21/05
3
Synopsys, Inc.
700 East Middlefield Road
Mountain View, CA 94043 USA
The CCS Noise analysis flow can take advantage of the advanced CCS
timing receiver model when such data are available in the library. More
details about the CCS timing receiver model can be found in the CCS
Timing White Paper [1]. For the purpose of noise analysis, the CCS
timing receiver model is more accurate than single-valued min/max
rise/fall pin capacitance because it includes the dependency of effective
receiver pin capacitance on input transition time and receiver output load
capacitance. Furthermore, the CCS Noise analysis engine explicitly
includes the varying effective input capacitance effect of victim receivers;
no extra receiver characterization is required.
Version 1.1
11/21/05
4
Synopsys, Inc.
700 East Middlefield Road
Mountain View, CA 94043 USA
Once characterized, CCS Noise model data are stored either on a timing
arc or on a pin in the cell library, depending on the topology of the circuit
netlist. For circuit cells having a single CCB for an input-output pin pair,
one CCS Noise model is extracted and stored on a timing arc. Such
single-stage cells include most inverters, NAND gates, NOR gates, AOI
gates, OAI gates, etc. For circuit cells having two subsequent CCBs, two
CCS Noise models are stored on a timing arc. Such two-stage cells
include most of the buffers, AND gates, OR gates, AND-OR gates, and
OR-AND gates, etc. For circuit cells having three or more CCBs,
including most of the flip-flops, full adders, macro blocks, etc., CCS
Noise model data are stored on pins.
Version 1.1
11/21/05
5
Synopsys, Inc.
700 East Middlefield Road
Mountain View, CA 94043 USA
injected noise
U1 U2 U3 FF
N1
Alternatively, you could fix just those noise bumps that can propagate all
the way to an endpoint such as the D pin of a flip-flop; this method of
analysis is referred to as report_at_endpoint mode.
injected noise
attenuated noise
U1 U2 U4 FF
N1
highly immune cell
11/21/05
6
Synopsys, Inc.
700 East Middlefield Road
Mountain View, CA 94043 USA
propagated noises
injected noise
U1 U2 U3 FF
N1
violation endpoint
violation source
Fig. 7. Source of violation vs. violation at an endpoint.
Version 1.1
11/21/05
7
Synopsys, Inc.
700 East Middlefield Road
Mountain View, CA 94043 USA
U1
U2 U4 FF
N1
U3
N2
backward searching for sources
Fig. 8 shows how the violation sources can be found given the violation
reported at the endpoint. The backward search starts from the endpoint,
and traces the fan-in logic cone until it hits the first net where the injected
noise exceeds the noise immunity. In this way, all possible sources of
violations can be found. In Figure 8, the violation has occurred at the D
pin of FF and report_at_endpoint mode reports nets N1 and N2 as the
violation sources. So when you fix the violations on N1 and N2, the
violation at the endpoint will be fixed as well.
Using CCS Noise and the actual shape of the input noise bump, the
noise immunity of the cell can be computed as well as the noise slack;
Version 1.1
11/21/05
8
Synopsys, Inc.
700 East Middlefield Road
Mountain View, CA 94043 USA
that is, the amount of noise height that needs to be added to the noise
bump to cause a failure. When the noise slack is negative, it means that
the noise bump has exceeded the noise immunity parameters for the cell.
When the noise slack is positive, the noise bump is within the noise
immunity parameters.
5 Results
Fig. 9 shows the results of a crosstalk noise correlation with SPICE. The
test circuit has two aggressor nets coupled to a victim net and the victim
net has a fan-out net for noise propagation analysis.
Fig. 9 (b) compares CCS Noise and SPICE waveforms for a typical case.
The difference in waveforms is barely discernable at the resolution of the
plot. The cell driving strengths: that is, the input transition times of the
aggressor drivers and the lengths of interconnects, have also been varied
to cover all reasonable scenarios for accuracy correlation.
The noise bump height correlation at point A (for noise calculation) is
shown in Fig. 9 (c). The correlation at point B (for both noise calculation
and propagation) is shown in Fig. 9 (d).
aggressor
victim A B
A B
Fig. 9. Comparison of CCS NoiseCCS NoiseCCS NoiseCCS Noise analysis results with SPICE.
Version 1.1
11/21/05
9
Synopsys, Inc.
700 East Middlefield Road
Mountain View, CA 94043 USA
Table 1 and Fig. 10 show details of static noise analysis results with
customer designs.
Version 1.1
11/21/05
10
Synopsys, Inc.
700 East Middlefield Road
Mountain View, CA 94043 USA
400
300
violations
200
100
0
Design A Design B Design C
6 Summary
In this paper we have introduced the new advanced CCS Noise model
and shown how it can be used to improve static noise analysis. CCS
Noise provides for accurate noise calculation including the effects of
noise propagation, so that the gate model can be characterized
significantly faster than before. Results with various designs show that
CCS Noise matches SPICE simulation results with excellent correlation.
7 References
[1] CCS Timing Technical White Paper.
Version 1.1
11/21/05
11